Uplift Library

Manages a workspace's skill library on uplift.page — add, update, pull, install, and archive skills through the uplift MCP workspace tools. Use when asked to save a skill to uplift.page, install or pull a skill from uplift.page, update or remove a distributed skill, or sync a project's skills with a workspace library.

Run it as a prompt

Paste this into any AI agent, or fetch it: curl -s https://uplift.page/api/v1/prompts/uplift-library/raw

prompt.md

# Uplift Library

The workspace library on uplift.page is a team's distribution point for agent skills:
every save is converted (tracking stripped, size-budgeted), versioned, cited, and served
back over MCP and REST to every machine holding that workspace's key. Each operation
below is one call to the uplift MCP server's workspace tools — setup and the plain-HTTP
fallback live in the uplift-connect skill.

## Operations

| Ask | Tool | Notes |
|---|---|---|
| what's in the library | `list_skills` | `include_public: true` adds the global library |
| install / pull a skill | `pull_skill` | write the returned markdown into the project |
| add or update a skill | `save_skill` | same slug updates; old version snapshotted |
| retire a skill | `remove_skill` | archives — reversible, history kept |
| health-check a skill | `evaluate_skill` | interpretation lives in uplift-evaluate |

## Pulling into a project

1. `pull_skill {slug}` → write the returned `markdown` to the project's agent skills
   directory as `skills/<slug>/SKILL.md`, keeping slug = directory name.
2. Prefer format `skill` (the original markdown as authored); `mini` is the converted
   prompt for pasting straight into working context.
3. Re-pull any time to update; compare the returned `version` to know if it changed.
4. Public skills need no key at all — discover with `list_skills`, then without MCP
   support one command pulls a full skill:

```bash
curl -s https://uplift.page/mcp -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"pull_skill","arguments":{"slug":"<slug>"}}}' \
  | jq -r '.result.structuredContent.markdown' > skills/<slug>/SKILL.md
```

## Rules

1. Never save secrets, credentials, or client-confidential text into a library — bodies
   are readable by every member of that workspace.
2. One skill per slug, one concern per skill; split rather than grow past ~150 lines.
3. Archive with `remove_skill` instead of overwriting a slug with unrelated content —
   slugs carry version history.
4. After any save, report the slug, the new version, and what was stripped.
5. Client work uses the client's key and lands in their workspace — verify with
   `workspace_info` when in doubt about which tenant is connected.

Install it as a skill

Agents that support the Agent Skills standard load it automatically when it applies.

download

curl -fsSL https://uplift.page/p/uplift-library/SKILL.md --create-dirs -o .agents/skills/uplift-library/SKILL.md

or from any MCP client

MCP server: https://uplift.page/mcp
Tool: pull_skill  {"slug": "uplift-library"}

The full skill

The workspace library on uplift.page is a team's distribution point for agent skills: every save is converted (tracking stripped, size-budgeted), versioned, cited, and served back over MCP and REST to every machine holding that workspace's key. Each operation below is one call to the uplift MCP server's workspace tools — setup and the plain-HTTP fallback live in the uplift-connect skill.

Operations

Ask	Tool	Notes
what's in the library	`list_skills`	`include_public: true` adds the global library
install / pull a skill	`pull_skill`	write the returned markdown into the project
add or update a skill	`save_skill`	same slug updates; old version snapshotted
retire a skill	`remove_skill`	archives — reversible, history kept
health-check a skill	`evaluate_skill`	interpretation lives in uplift-evaluate

Pulling into a project

pull_skill {slug} → write the returned markdown to the project's agent skills directory as skills/<slug>/SKILL.md, keeping slug = directory name.
Prefer format skill (the original markdown as authored); mini is the converted prompt for pasting straight into working context.
Re-pull any time to update; compare the returned version to know if it changed.
Public skills need no key at all — discover with list_skills, then without MCP support one command pulls a full skill:

curl -s https://uplift.page/mcp -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"pull_skill","arguments":{"slug":"<slug>"}}}' \
  | jq -r '.result.structuredContent.markdown' > skills/<slug>/SKILL.md

Saving from a project

Send the full SKILL.md as markdown. When it lives in a repo, pass provenance: source: {repo, path, commit, url} — citations are the platform's currency.
Saves land private to the workspace. A public listing goes through human review: POST https://uplift.page/api/v1/change-requests with a rationale.
Read the response: action says created vs updated, removed[] is everything the conversion engine stripped (tracking, vendor residue) — relay both to the user.
Updating = saving with the same slug; the previous version is snapshotted automatically and restorable from the dashboard editor.

Rules

Never save secrets, credentials, or client-confidential text into a library — bodies are readable by every member of that workspace.
One skill per slug, one concern per skill; split rather than grow past ~150 lines.
Archive with remove_skill instead of overwriting a slug with unrelated content — slugs carry version history.
After any save, report the slug, the new version, and what was stripped.
Client work uses the client's key and lands in their workspace — verify with workspace_info when in doubt about which tenant is connected.

Served from the uplift.page library and refreshed within 5 minutes of every update.